Cheem tour; 1D manual tour on the selected attribution

Create a linked plotlyof the global data- and attribution- spaces. Typically consumed directly by shiny app.

radial_cheem_tour(
  cheem_ls,
  basis,
  manip_var,
  primary_obs = NULL,
  comparison_obs = NULL,
  do_add_pcp_segments = TRUE,
  pcp_shape = c(142, 124, 3),
  angle = 0.15,
  row_index = NULL,
  inc_var_nms = NULL,
  do_center_frame = TRUE,
  do_add_residual = FALSE
)

radial_cheem_tour_subplots(
  cheem_ls,
  basis,
  manip_var,
  primary_obs = NULL,
  comparison_obs = NULL,
  do_add_pcp_segments = TRUE,
  pcp_shape = c(142, 124, 3),
  angle = 0.15,
  row_index = NULL,
  inc_var_nms = NULL,
  do_center_frame = TRUE
)

Arguments

cheem_ls

A return from cheem_ls(), a list of data frames.

basis

A 1D projection basis, typically a return of basis_attr_df().

manip_var

The , number of the manipulation variable.

primary_obs

The rownumber of the primary observation. Its local attribution becomes the 1d projection basis, and the point it highlighted as a dashed line. Defaults to NULL, no primary observation highlighted.

comparison_obs

The rownumber of the comparison observation. Point is highlighted as a dotted line. Defaults to NULL, no comparison observation highlighted.

do_add_pcp_segments

Logical, whether or not to add parallel coordinate line segments to the basis display.

pcp_shape

The number of the shape character to add. Expects 142, 124, 3 '|' for plotly and gganimate or '+' in either respectively. Defaults to 142, '|' for plotly.

angle

The step size between interpolated frames, in radians. Defaults to .15.

row_index

Numeric index of selected observations. Defaults to TRUE; 1:n.

inc_var_nms

A vector of the names of the variables to include in the projection.

do_add_residual

Whether of not to add a facet with a fixed y on residual. Doing so may cause issues with animation. Defaults to FALSE.

do_scale_in_frame

Whether or not to scale by standard deviations away from the mean within each frame or not. Defaults to TRUE, helping to keep the animation centered.

Value

ggtour (ggplot2 object with frame info) animation frames of a radial tour manipulating the contribution of a selected tour. Animated with spinifex::animate_* functions.

See also

Other cheem consumers: global_view(), run_app()

Examples

library(cheem)
library(spinifex)

## Classification:
X    <- penguins_na.rm[, 1:4]
clas <- penguins_na.rm$species
Y    <- as.integer(clas)

rf_fit  <- default_rf(X, Y)
## Long runtime for full datasets or complex models:
shap_df <- attr_df_treeshap(rf_fit, X, noisy = FALSE)
this_ls <- cheem_ls(X, Y, class = clas,
                     model = rf_fit,
                     attr_df = shap_df)

bas <- basis_attr_df(shap_df, rownum = 1)
ggt <- radial_cheem_tour(this_ls, basis = bas, manip_var = 1,
  primary_obs = 1, comparison_obs = 2)
if (FALSE) {
animate_plotly(ggt)
if(FALSE) ## or animate with gganimate
  animate_gganimate(ggt) #, render = gganimate::av_renderer())

## Regression:
dat  <- amesHousing2018_NorthAmes
X    <- dat[, 1:9]
Y    <- log(dat$SalePrice)
clas <- dat$SubclassMS

rf_fit  <- default_rf(X, Y)
## Long runtime for full datasets or complex models:
shap_df <- attr_df_treeshap(rf_fit, X, noisy = FALSE)
this_ls <- cheem_ls(X, Y, class = clas,
                     model = rf_fit,
                     attr_df = shap_df)

bas <- basis_attr_df(shap_df, rownum = 1)
ggt <- radial_cheem_tour(this_ls, basis = bas, manip_var = 1)

animate_plotly(ggt)
if(FALSE) ## or animate with gganimate
  animate_gganimate(ggt, render = gganimate::av_renderer())
}


## Experimental radial tour made from plotly::subplots rather than facets
ggt <- radial_cheem_tour_subplots(this_ls, basis = bas, manip_var = 1)
animate_plotly(ggt)